In lipid membranes, PLN adopts an L-shaped conformation, with a membrane-adsorbed, amphipathic regulatory region (domain Ia, M1 to T17) connected by a short loop (Ile18 to Gln22) to a helical inhibitory region (domains Ib, Gln23 to Asn30; and domain II, Leu31 to Leu52), which crosses the SR membrane (Traaseth et al., 2009; Verardi et al., 2011). In its storage form, PLN is pentameric (Verardi et al., 2011; Vostrikov et al., 2013; Mravic et al., 2019) and de-oligomerizes into active L-shape monomers (Traaseth et al., 2009). The dynamic cytoplasmic region undergoes an order-disorder transition between tense (T) and relaxed (R) states, with the latter promoted by Ser16 phosphorylation (Seidel et al., 2008; Gustavsson et al., 2013; Gustavsson et al., 2012). Upon binding SERCA, domain Ia transitions to a more rigid and non-inhibitory bound (B) state, becoming more populated upon phosphorylation (Seidel et al., 2008; Gustavsson et al., 2013). How does Ser16 phosphorylation signal the reversal of inhibition to the TM region? Since the inhibitory TM region is ~45 Å away from Ser16 and ~20 Å from SERCA’s Ca²⁺-binding sites, we speculated that both phosphorylation (of PLN) and Ca²⁺ binding (to SERCA) must transmit conformational and topological changes across the membrane, thus allosterically modulating SERCA's function.

Residue-specific anisotropic NMR parameters such as CSs and DCs are exquisitely suited to describe topological transitions such as tilt, bend, and torque of TM proteins in lipid bilayers near-physiological conditions (Opella and Marassi, 2004; Das et al., 2013). Their analysis by OS-ssNMR requires that membrane-embedded proteins are uniformly oriented relative to the static magnetic field (B₀). Therefore, we reconstituted PLN free and in complex with SERCA into magnetically aligned lipid bicelles (Sanders and Landis, 1995). Since monomeric PLN is the functional form (Zvaritch et al., 2000), we utilized a cysteine-null monomeric mutant of PLN (Cys36Ala, Cys41Phe, and Cys46Ala) (Karim et al., 2000) exclusively throughout this study. Both unphosphorylated and phosphorylated (pPLN) variants of PLN were expressed recombinantly, while SERCA was purified from mammalian tissues (Stokes and Green, 1990). Since lipid bicelles orient spontaneously with the normal of the membrane (n→) perpendicular to B₀, we doped the sample with Yb³⁺ ions to change the magnetic susceptibility and orient the lipid membranes with n→ parallel to B₀. This expedient doubles the values of CSs and DCs and increases the resolution of the NMR spectra (Prosser et al., 1996; Prosser et al., 1998). Figure 1A shows the 2D [¹⁵N-¹H] sensitivity-enhanced (SE)-SAMPI4 (Gopinath and Veglia, 2009) separated local field (SLF) spectra of free PLN and pPLN. Due to PLN's intrinsic conformational dynamics, the SLF spectra visualize only its TM region. The spectra display the typical wheel-like pattern diagnostic of a helical conformation for both the TM domains of PLN and pPLN. Residue-specific assignments were carried out on free PLN using a combination of a 3D SE-SAMPI4-proton-driven spin diffusion (PDSD) spectrum (Mote et al., 2011), selective ¹⁵N-labeled samples, and predictions from MD simulations (Weber and Veglia, 2020; Table 1; Figure 1—figure supplements 1–3). To obtain PLN’s topology in lipid bilayer, the assigned resonances were fit to idealized Polar Index Slant Angle (PISA) models extracting whole-body tilt (θ) and rotation or azimuthal (ρ) angles (Denny et al., 2001; Marassi and Opella, 2000), which for free PLN was θ = 37.5 ± 0.7° and ρ_L31 = 201 ± 4°, and for pPLN θ = 34.8 ± 0.5° and ρ_L31 = 201 ± 4°, where ρ_L31 is the rotation angle referenced to Leu31. Notably, the high resolution of the oriented SLF spectra of PLN show two distinct sets of peaks (Figure 1B), with populations unevenly distributed. The average population of the minor state estimated from the normalized peak intensities is approximately 30 ± 9 %. Remarkably, the resonances of the minor population overlap almost entirely with those of pPLN (Figure 1B), revealing a topological equilibrium in which the TM region of PLN interconverts between two energetically different orientations. We previously showed that PLN phosphorylation shifts the conformational equilibrium toward the R state (Gustavsson et al., 2013), releasing the interactions with the lipid membranes of domain Ia (Figure 1C). Our OS-ssNMR data show that these phosphorylation-induced effects propagate to the TM domains, shifting the topological equilibrium toward the less populated state.

To investigate how Ser16 phosphorylation allosterically affects the inhibitory TM binding interface, we reconstituted the SERCA/PLN complex in lipid bicelles and studied it by OS-ssNMR. The alignment of mammalian SERCA in bicelles was confirmed by cross-linking the most reactive cysteines with a trifluoromethylbenzyl (TFMB)-methanethiosulfonate (MTS) tag to probe its alignment by ¹⁹F NMR (Figure 1D; Figure 1—figure supplement 4). Five of the 24 cysteines of SERCA were uniquely labeled as monitored by solution NMR in isotropic bicelles (q = 0.5), with two positions being the most prominent. In anisotropic bicelles (q = 4.0) and at low temperature, the ssNMR spectrum of ¹⁹F-SERCA consists of a single unresolved ¹⁹F resonance due to the rapid reorientation of the enzyme in the isotropic phase. Upon increasing the temperature, the ¹⁹F-SERCA/bicelle complex orients with n→ perpendicular to B₀, and the ¹⁹F resonance becomes anisotropic as a triplet with 1.3 kHz dipolar coupling and no indication of any powder pattern associated with non-aligned SERCA (Glaser et al., 2004). Figure 1E shows the 2D SLF spectra of PLN and pPLN in complex with SERCA. To maintain a functional and stable complex, we used a lipid-to-complex molar ratio of 2000:1, equivalent to one SERCA per bicelle at this q-ratio (Glover et al., 2001), with PLN concentration 10 times less than in the SERCA-free samples. Therefore, the signal-to-noise ratio in the oriented spectra is significantly reduced relative to the free forms. Nonetheless, the SLF spectra of both SERCA/PLN and SERCA/pPLN complexes show the wheel-like patterns typical of the α-helical domains with selective exchange broadening for resonances located at the protein-protein binding interface (Figure 1F). The assigned peaks associated with the helical domain II were fit to the ideal PISA model, yielding θ = 33.2 ± 1.2° and ρ_L31 = 193 ± 7°. Therefore, upon binding SERCA, PLN requires a distinguishable -4.3 ± 1.4° change in tilt (p = 4.3 × 10⁻¹⁵) and a less significant -8 ± 8° change in rotation (p = 0.00027). These error bounds factor the linewidths and variation associated with substituting ambiguous assignments into the PISA fitting (parentheses of Figure 1E). Similarly, the PISA model for pPLN was fit to θ = 30.4 ± 1.1° and ρ_L31 of 197 ± 4°, suggesting that the topology of the TM domain requires adjustments of -3.4 ± 1.2° (p = 2.9 × 10⁻¹⁴) and -4 ± 6° (p = 0.0013) to form a complex with the ATPase.

Reductions in the TM helix tilt angle, which accompanied phosphorylation and complex formation, also coincided with a dramatic broadening of peaks in the cluster of isotropic resonances around 140 ppm (Figure 1E). These resonances are attributed to the dynamic domain Ib residues, and their disappearance is consistent with SERCA binding, which requires the unwinding of the juxtamembrane region, and a concomitant reduction of the tilt angle to re-establish hydrophobic matching with the thickness of the lipid bilayer (Karim et al., 2006; Gustavsson et al., 2013). Therefore, for free and bound PLN, we find that communication between cytoplasmic and intramembrane environments is transduced via domain Ib dynamics. Analysis of the SLF spectra also shows that phosphorylation of PLN at Ser16 restores the intensities of most resonances except for those at the upper binding interface (i.e. Asn30, Leu31, Asn34, and Phe41). These spectral changes suggest a reorganization of PLN-SERCA packing interactions, rather than a complete dissociation of the complex, consistent with prior MAS-ssNMR, EPR, and FRET measurements (Karim et al., 2006; Gustavsson et al., 2013; Dong and Thomas, 2014; Bidwell et al., 2011).

To better understand how PLN phosphorylation relieves SERCA inhibition, we determined the structural ensembles of the SERCA/PLN and SERCA/pPLN complexes in the non-inhibitory bound (B) state by incorporating the data from our experimental measurements into RAOR-MD samplings (De Simone et al., 2014; Sanz-Hernández et al., 2016). This dynamic refinement methodology employs full atomic MD simulations in explicit lipid membranes and water and utilizes restraints from sparse datasets to generate experimentally driven structural ensembles. As starting coordinates for our samplings, we used the X-ray structure of E2-SERCA/PLN, where a super-inhibitory mutant of PLN was used to stabilize the complex for crystallization (Akin et al., 2013). We docked the TM domains of PLN using restraints obtained from chemical cross-linking experiments for both cytoplasmic and luminal sites (Toyoshima et al., 2003; Chen et al., 2006; Chen et al., 2003; Figure 2—figure supplement 1A). The dynamic cytoplasmic region (loop and domain Ia), which was not resolved in the crystal structure, was constrained to a region between the nucleotide-binding (N) and phosphorylation (P) domains of SERCA via paramagnetic relaxation enhancements (PRE) obtained from MAS-ssNMR of the complex in the B-state (Gustavsson et al., 2013). Chemical shifts have indicated that PLN’s domain Ia binds SERCA’s cytoplasmic headpiece in an extended conformation (Gustavsson et al., 2013). Intermolecular PREs included quenching of PLN domain Ia resonances by SERCA bearing a paramagnetic (1-oxyl-2,2,5,5-tetramethylpyrroline-3-methyl) methanethiosulfonate spin label (MTSSL) on Cys674 and quenching of ¹³C-methyl thiocysteines (MTC), engineered throughout SERCA’s N and P domains, by MTSSL-labeled PLN (Figure 2—figure supplement 1A). Due to the dynamic nature of domain Ia and the ambiguity of ¹³C-MTC assignments, all PRE restraints were implemented as 30 Å boundary conditions to confine the conformational freedom of PLN domain Ia to the approximate binding site (Figure 2—figure supplement 1B). Using this scheme, the overall profile of the average pairwise distances for residues in domain Ia and loop to the spin-label at Cys674 matches the PRE measurements, with the minimal distance (i.e. maximum PRE effect) observed for PLN-Tyr6 (Figure 2—figure supplement 1C,D).

CSs and DCs from OS-ssNMR were applied to the TM region as ensemble-averaged restraints across eight replicas. Back-calculated CS and DC values for PLN and pPLN were in excellent agreement with experiments (Figure 2—figure supplement 1E). Average back-calculated tilt angles of 32.8° and 30.4° for PLN and pPLN, respectively, matched PISA fits to experimental values (Figure 2—figure supplements 1F and 2). All pairwise distances between previously reported cross-linkable positions (Toyoshima et al., 2003; Chen et al., 2006; Chen et al., 2003; Jones et al., 2002) were distributed within acceptable ranges (Figure 2—figure supplement 1G). Convergence of Q-factors of DC and DC restraints during the annealing cycles is shown in Figure 2—figure supplement 3A. Although not used as an initial docking restraint, cytoplasmic residues PLN Lys3 and SERCA Lys397 were also partially distributed within a distance consistent with previously reported cross-linking (James et al., 1989). Overall, the resulting structural ensembles were in excellent agreement with all the available experimental data for both complexes.

While TM and cytoplasmic regions of PLN were restrained throughout the simulations, SERCA was unrestrained and displayed a significant conformational heterogeneity over the timescale sampled. To assess SERCA’s conformational landscapes, we used principal component analysis (PCA), clustering the ensembles according to two key motions, exemplifying a combined opening of the cytoplasmic headpiece involving a hinge-like displacement of the N domain and rotation of the A domain away from the P domain (PC1) and planar rotations separating the N and A domains (PC2) (Figure 2A,B; Figure 2—figure supplement 4, Figure 2—videos 1, 2). These motions differentiate the E1 and E2 states of SERCA, as shown by the projections of crystal structures onto the PCA map (Figure 2C,D). The SERCA/PLN complex spans four distinct clusters, while the SERCA/pPLN complex spans eight (see Figure 2—figure supplement 5 for representative structures). Both ensembles converged with respect to both PC1 and PC2 motions based on the Kullback-Leibler divergence (KLD) plot of the 2D PCA histograms. Note that the convergence of the SERCA/pPLN ensemble was slower due to the additional conformational dynamics (Figure 2—figure supplement 3B). When bound to PLN, SERCA mostly retains the compact E1-like headpiece present in the crystal structure and predominantly occupied a highly compact cluster (3) resembling nucleotide-bound E1 states and a cluster (1) intermediate toward the E2 states. Albeit biased along the E1 coordinate of PC2, similar states were present for pPLN, but the interaction of pSer16 with Arg604 weakens the Asp601-Thr357 and Arg604-Leu356 hydrogen bonds at the hinge of the N and P domains, leading to four additional open states (clusters 5–8). Separate clusters correspond to successive breakages of interdomain hydrogen bonds in the headpiece. Salt bridges between PLN-Ser16 to SERCA-Arg460 and PLN-Arg14 to SERCA-Glu392 were also found to stabilize these additional open states (Figure 2—figure supplement 6). These open states resemble off-pathway crystal structures solved for the Ca₂E1 state observed in the absence of nucleotide (Dyla et al., 2020; Toyoshima et al., 2000; Jensen et al., 2006). For all clusters, the binding interactions near the Ser16 position were more persistent for pPLN than PLN (Figure 2E,F; Figure 2—figure supplement 5). Although our experimental data cannot verify these conformational landscapes of SERCA, our ensembles nonetheless strongly suggest that phosphorylation of PLN has long-range effects on the complex well beyond the protein-protein binding interface.

From the analysis of the structural ensembles of the two complexes, it emerges that the relief of inhibition (i.e. activation) occurs via a rearrangement of the intramembrane contacts between the TM region of pPLN and SERCA, with a reconfiguration of electrostatic interactions near the phosphorylation site and a disruption of packing at the protein-protein interface (Figure 3A,B). The interactions between the cytoplasmic regions are transient and highly dynamic, resembling the conformational ensembles of intrinsically disordered complexes (Olivieri et al., 2020). For both complexes, we observe persistent interactions between PLN-Glu2 and SERCA-Lys365, PLN-Glu19, and SERCA-Lys328, and to a lesser extent PLN-Lys3 and SERCA-Asp399 and PLN-Tyr6 and SERCA-Asp557. For PLN, however, the arginine residues (Arg9, Arg13, and Arg14) interact transiently with SERCA-Glu606, while for pPLN, the phosphate group at Ser16 interacts strongly with SERCA-Arg604 and SERCA-Lys605. Residual interactions observed for PLN residues Arg13, Arg14, Ser16, and Thr17 with lipid headgroups were abolished with pPLN (Figure 3A), consistent with the correlation between membrane-detachment of domain Ia and reduced inhibition (Gustavsson et al., 2012; Gustavsson et al., 2011). Interestingly, we detected the formation of intramolecular salt-bridges between the phosphate of Ser16 and PLN-Arg9, PLN-Arg13, and PLN-Arg14, causing domain Ia to adopt a compact conformation as previously suggested by fluorescence data (Li et al., 2003; Li et al., 2004; Tables 2 and 3). These cytoplasmic protein-protein interactions destabilized PLN’s domain Ib and consequently severed inhibitory intermolecular contacts with SERCA’s TM helices involving Gln23-Leu321/Arg325 (M4), Lys27-Phe809 (M6), and Asn30-Trp107 (M2), while the intermolecular contacts involving domain II are mostly retained (Figure 3A and Figure 3—figure supplement 1). For pPLN, the disrupted domain Ib-SERCA interactions were substituted by the interactions with the lipid headgroup, suggesting that the membrane itself may also play a localized role in modulating inhibition (Figure 3A). In fact, hydrophobic substitutions within domain Ib have been identified as hotspots for engineering super-inhibitory PLN mutants (i.e. Asn/Lys27Ala and Asn30Cys), which exhibit stable helical structure well into the loop domain (Akin et al., 2013; Kimura et al., 1998; Akin et al., 2010). Importantly, these structural ensembles capture the order-disorder dynamics of domain Ib resonances observed in the OS-ssNMR spectra and suggested by previous NMR and EPR studies (Gustavsson et al., 2011). Destabilization and detachment of this region is consistent with the reappearance of exchange-broadened interfacial resonances of domain II paralleled by broadening of the resonances of the dynamic domain Ib in the SLF spectra of the SERCA/pPLN complex (Figure 1E). The interactions of domain Ia (Arg13-Ser16) and inhibitory contacts of domain Ib (at Lys27) with SERCA appear to be mutually exclusive (Figure 3C–F). In both complexes, the electrostatic interactions of PLN-Arg13, PLN-Arg14, or PLN-pSer16 with SERCA’s Arg604-Glu606 stretch cause the detachment of PLN’s domain Ib and the consequent weakening of the inhibitory interaction (Figure 3C–G). This illustrates the regulatory role of the B state of PLN for relieving inhibition and the super-inhibitory activity of domain Ia-truncated PLN (Gustavsson et al., 2013).

To assess the effects of PLN’s phosphorylation on Ca²⁺ transport, we calculated the topological correlations of the TM’s tilt angle fluctuations between PLN's and SERCA’s TM helices (Figure 4A–D). When PLN is bound to SERCA, we observe a dense network of correlated motions between PLN's TM region and the binding groove (TM2, TM6, and TM9), as well as a dense cluster of correlations involving TM3, TM4, TM5, TM6, and TM7. This allosteric coupling influences the Ca²⁺-binding sites' geometry, possibly reducing SERCA's Ca²⁺ binding affinity. In contrast, the analysis of the trajectories of the SERCA/pPLN complex displays only correlated motions between the TM of pPLN and the most proximal SERCA helices, with only a sparse network of correlations involving TM4, TM5, TM6, and TM8. Phosphorylation of PLN at Ser16 increases the electrostatic interactions with the cytoplasmic domain of SERCA (R to B state transition, i.e. disorder to order) (Gustavsson et al., 2013), and simultaneously weakens intramembrane protein-protein interactions to uncouple the dynamic transitions of PLN from SERCA. The latter removes the structural hindrance of PLN’s TM domain and augments Ca²⁺ transport (Figure 4E,F).

To assess the effects of Ca²⁺, we performed SLF experiments on SERCA/PLN and SERCA/pPLN complexes in the E1 state (Figure 5A,B). The addition of Ca²⁺ to the SERCA/PLN complex did not cause significant changes to the PLN topology (θ = 32.9 ± 1.4° and ρ_L31 = 199 ± 4°; p=0.29 and p=0.0025 compared to the E2 complex). However, the reappearance (i.e. sharpening) of several resonances in the spectra (e.g. Phe32, Phe35/Leu42, and Ala36) indicates a rearrangement of the binding interface between the two proteins similar to phosphorylation's effect on the E2 complex. A small topological change, however, was observed for the SERCA/pPLN complex, for which Ca²⁺-binding induced a decrease of both tilt and rotational angles of 1.8 ± 1.3° and 10 ± 7°, respectively (θ = 28.6 ± 0.7° and ρ_L31 = 187 ± 6°, p=1.9 × 10⁻⁶ and p=4.2 × 10⁻⁷ compared to the E2 complex). Due to the lack of X-ray structures, we were unable to carry out dynamic modeling of these complexes. However, these experimental results agree with our E2-SERCA models suggesting that a loss of inhibition, either from phosphorylation or Ca²⁺ binding, does not require an extensive structural and topological reconfiguration of PLN’s domain II or complete dissociation of the complex.

Expression plasmids used in this study are available on request.

All experiments were caried out in vitro using PLN obtained in this work by recombinant expression. Recombinant PLN was expressed in E. coli BL21 (DE3) cells grown in minimal media required for uniform or selective ¹⁵N labeling. SERCA1a was purified from crude ER stored at −80°C. The crude ER was prepared from skeletal muscle harvested from New Zealand white rabbits immediately following euthanasia (approved IACUC Protocol 1805-35910A).

Long-chain lipids DMPC (37.0 mg), POPC (10.4 mg) and PE-DTPA (0.9 mg; that is, 79.25:19.75:1.0 molar ratio) were aliquoted together from chloroform stocks (Avanti Polar Lipids), dried to a film with N₂ and residual solvent removed under high vacuum. The film was resuspended into 1 mL ddH₂O, freeze-thawed three times between liquid N₂ and a 40°C water bath then lyophilized. DHPC (7.7 mg; 1:4 molar ratio, or q = 4, to long chain lipids) was prepared separately from a chloroform stock, dried and lyophilized from ddH₂O.

For bicelles containing only PLN, DHPC was dissolved into 250 μL sample buffer (20 mM HEPES, 100 mM KCl, 1 mM NaN₃, 2.5% glycerol, pH 7.0), then used to solubilize PLN (2.5 mg) by vortex. Separately, long-chain lipids were suspended into 250 μL of buffer. PLN in DHPC and long-chain lipids, both pre-chilled in ice, were combined and vortexed while allowing the sample to reach room temperature, then placed back on ice. The process was repeated at least three times to fully solubilize long-chain lipids, which produced a completely transparent liquid at cold temperature (micelle phase) and transparent solid gel at room temperature (bicelle phase). The sample was placed on ice, brought to pH 4.2 with KOH, then concentrated to ~180 μL using a 0.5 mL 10 kDa MWCO centrifugal filter (Amicon) at 4°C. The solubility of PLN and pPLN was significantly diminished at higher pH. The sample was doped with 0.8 μL of 1 M YbCl₃, corrected back to pH 4.2 with KOH, then loaded into a 5 mm flat bottom sample cell (New Era).

For bicelles containing the SERCA/PLN complexes, SERCA1a was purified from rabbit skeletal muscle as previously described (Stokes and Green, 1990). SERCA was eluted from Reactive Red affinity resin at ~0.5 mg/mL in SERCA, 0.1% C₁₂E₈, 1 mM CaCl₂, 1 mM MgCl₂, 20 mM MOPS, 20% glycerol, 8 mM ADP, 0.25 mM DTT, pH 7.00 and stored at −80°C. Protein concentration was determined by Pierce BCA Assay (Thermo Scientific) and activity confirmed by coupled-enzyme assays (Reddy et al., 2003). Immediately prior to use, SERCA (4 mg) was thawed at 4°C and combined with long-chain lipids (prepared as above, except with 1.5% molar PE-DTPA chelating lipid) solubilized into 1 mL of 4% C₁₂E₈ in sample buffer. The mixture was diluted to ~30 mL with sample buffer and stirred at 4°C for 30 min prior removal of C₁₂E₈ by adding 4 g of Bio-Beads SM-2 (Bio-Rad) in stages of 0.5, 0.5, 1, and 2 g with 15 min stirring between additions. Stirring continued overnight at 4°C. Bio-Beads were removed by a 25G syringe and the cloudy suspension of proteoliposomes centrifuged at (12,000 rpm, JA25.50 rotor, 4°C, 30 min). The pellet was resuspended into ~40 mL sample buffer and centrifuged once more to wash out residual elution buffer. The final pellet was resuspended with 250 μL sample buffer and fully solubilized by adding a 250 μL mixture of PLN, or pPLN (0.23 mg), in DHPC (adjusted to pH 7.0) with several cooling/heating cycles under vortex. The bicelle mixture (~1 mL) was centrifuged (13,400 rpm, Eppendorf F45-12-11 rotor, 4°C, 30 s) to remove insoluble debris and the supernatant concentrated to ~200 μL using a 0.5 mL 10 kDa MWCO centrifugal filter (Amicon) at 4°C. The sample was doped with 1.6 μL of 1 M YbCl₃ in four stages, correcting pH back to 7.0 with KOH at each addition. Sample buffer included 20 mM HEPES, 100 mM KCl, 1 mM NaN₃, 5 mM MgCl₂, 2 mM DTT, 2.5% glycerol, pH 7.0 with 4 mM EGTA or 5 mM CaCl₂ to stabilize E2 or E1 states, respectively. SDS-PAGE confirmed co-reconstitution of SERCA and PLN in the bicelles (Figure 7D).

Trifluoromethylbenzyl (TFMB)-methanethiosulfonate (MTS) was synthesized analogously to our method previously reported for synthesizing a ¹³C-ethylmethanethiosulfonate reagent (Vostrikov et al., 2016; Weber et al., 2019) with minor modifications. Briefly, 4-(trifluoromethyl)benzyl bromide (5 mmole), MTS (5 mmole) and KI (0.03 mmole) were dissolved into 2 mL of dimethylformamide and stirred under nitrogen for 16 hr at 40°C.

For TFMB tagging, SERCA (8 mg) was thawed and dialyzed into 1 L of E1-state sample buffer (as above, but without DDT and including 0.25 mM C₁₂E₈) overnight at 4°C. DMPC (25.2 mg) and POPC (7.1 mg), solubilized in 500 μL of 5% C₁₂E₈, was then added to dialyzed SERCA. Detergent was removed by stirring with 1 g Bio-Beads SM-2 for 1.5 hr at 4°C and a further 30 min at room temperature. Proteoliposomes were removed from Bio-Beads with a 25G syringe and centrifuged (18,000 rpm, JA25.50 rotor, 4°C, 40 min). The pellet was suspended into 120 μL buffer and solubilized by adding DHPC (q = 4 for oriented or q = 0.5 for isotropic bicelles). TFMB-MTS (200 mM in DMSO) was added to the bicelles at 5:1 molar excess and incubated at room temperature for 1 hr prior to concentrating to ~250 with a 0.5 mL 10 kDa MWCO centrifugal filter (Amicon) and loading into a Shigemi NMR tube for ¹⁹F NMR measurement.

¹⁹F NMR spectra of TFMB-tagged SERCA were acquired on a solution state Bruker 600 MHz Avance NEO spectrometer equipped with a TCI HCN cryoprobe. 1D single-pulse experiments were acquired using a 90° pulse of 12 µs and recycle delay of 0.4 s. Spectra in isotropic bicelles were acquired with 1 k scans and 4 k scans for oriented bicelles. Spectra were processed using NMRPipe (Delaglio et al., 1995).

All ¹⁵N spectra were acquired on a Varian VNMRS spectrometer equipped with a low-E static bicelle probe (Gor'kov et al., 2007) operating at a ¹H frequency of 700 MHz. 1D [¹H-¹⁵N]-cross-polarization (CP)-based experiments used 90° pulse length of 5 μs, or 50 kHz radiofrequency (RF) field, on ¹H and ¹⁵N channels; contact time of 500 μs with a 10% linear ramp on ¹H centered at 50 kHz; and an acquisition time of 10 ms under 50 kHz SPINAL64 heteronuclear proton decoupling (Fung et al., 2000). For all experiments, ¹⁵N was set to 166.3 ppm and externally referenced to ¹⁵NH₄Cl at 39.3 ppm (Bertani et al., 2014); and detected using a spectral width of 100 kHz.

2D separated local field (SLF) spectra were collected using a signal-enhanced (SE)-SAMPI4 experiment (Gopinath and Veglia, 2009; Gopinath et al., 2010; Nevzorov and Opella, 2007). The indirect dipolar dimension utilized complex points and a spectral width of 31.25 kHz. The t₁ evolution period utilized ¹H homonuclear decoupling with an RF field of 50 kHz and 48 μs dwell time, and a phase-switched spin-lock pulses on ¹H and ¹⁵N of 62.5 kHz RF field. The sensitivity enhancement block used a τ delay of 75 μs and three cycles of phase-modulated Lee-Goldberg (PMLG) homonuclear decoupling (Vinogradov et al., 1999) with an effective RF field of 80 kHz.

3D SE-SAMPI4-PDSD spectra (Mote et al., 2011) were acquired with 15 increments in both indirect dimensions; spectral widths of 31.25 kHz and 10 kHz in the dipolar coupling and indirect ¹⁵N dimensions, respectively; and 3 s mixing time for ¹⁵N-¹⁵N diffusion (Traaseth et al., 2010). Total acquisition times for PLN samples were typically 1 hr for a 1D [¹H-¹⁵N] CP spectra (1 k scans), 40 hr for 2D SE-SAMPI4 spectra (1 k scans) and 2 weeks for 3D SE-SAMPI4-PDSD spectra (two experiments added with 0.25 k scans each). For the SERCA/PLN complexes, 2D SE-SAMPI4 spectra were acquired over 4.5 days at 25°C (4 k scans, 15 indirect points) and 1D [¹H-¹⁵N] CP spectra for 4 hr (4 k scans). A recycle delay of 3 s was used for all experiments. All spectra were processed using NMRPipe (Delaglio et al., 1995) and analyzed using NMRFAM-SPARKY (Lee et al., 2015) and Nmrglue (Helmus and Jaroniec, 2013).

Polar Index Slant Angle (PISA) wheels (Denny et al., 2001; Marassi and Opella, 2000) were fitted to 2D SE-SAMPI4 spectra using the PISA-SPARKY plugin in the NMRFAM-SPARKY package (Weber et al., 2020a). Default parameters were used to describe ideal helix structure and ¹⁵N chemical shift (CS) and ¹⁵N-¹H dipolar coupling (DC) tensors. The Cα-N-H bond angle was modified from 116° to 119° to best fit PLN spectra. CSs and DCs were fit by exhaustively searching tilts (θ), rotations (ρ_L31, i.e. referenced to residue Leu31) and order parameters (S) in increments of 0.1°, 1.0°, and 0.01, respectively, for the lowest RMSD between calculated and experimental values (Weber et al., 2020a). The parameter S, which factors scaling due to rigid-body helical fluctuations and imperfect alignment (Weber et al., 2020b), was determined as 0.80 ± 0.05 and 0.81 ± 0.03 for PLN and pPLN, respectively, alone in bicelles. Due to the sparsity of peaks assigned for spectra of PLN in complex with SERCA, S was fixed to 0.80 to reduce fitting errors. Errors in tilt θ, ρ_L31 and S were determined by repeating fitting 20 times with peak positions randomly adjusted at each iteration. Random adjustments were taken from a Gaussian distribution having a standard deviation equal to average FWHM peak linewidths (3 ppm for CS and 0.8 kHz for DC dimensions). The statistical significance of topological comparisons made throughout the text were determined using an unpaired t-test (Microsoft Excel 365 Version 2002) on tilt and azimuthal angles from the 20 repeat fits.

A simulation of truncated monomeric PLN (Met20 to Leu52; PDB 2LPF De Simone et al., 2013) in 97 DMPC and 32 POPC, 150 mM KCl and 4927 waters was constructed using the CHARMM-GUI webserver (Wu et al., 2014; Jo et al., 2008) and CHARMM36 forcefield (Huang and MacKerell, 2013). Production runs were done using the AMBER18 (Case et al., 2018; Crowley et al., 2009) package at 10 Å cutoff and 8 Å force-based switching and default configuration files provided by the CHARMM-GUI webserver. For example, using the Langevin thermostat (Loncharich et al., 1992) (310 K), Monte Carlo barostat (Faller and de Pablo, 2002) (1 bar, semi-isotropic coupling) and the SHAKE algorithm (Miyamoto and Kollman, 1992) for constraining hydrogens. The simulation was run for 1 µs. The final 900 ns of trajectory was used to predict ¹⁵N chemical shifts and ¹⁵N-¹H dipolar couplings, as previously reported (Weber and Veglia, 2020).

The initial conformation of SERCA in the E2 state was obtained from the crystal structure of the SERCA/PLN complex (PDB 4Y3U) (Akin et al., 2013). Missing loops in the structure of SERCA were introduced using MODELLER software (Webb and Sali, 2016). Since the super-inhibitory mutant of PLN used for crystallization harbors native cysteines and four substitutions not present in the monomeric mutant used in NMR experiments, as well as a missing C-terminal helical turn, the interface between SERCA and monomeric PLN was refined in silico by introducing information from cross-linking data. Specifically, the transmembrane section of PLN was docked onto SERCA in vacuo, maintaining backbone positional restraints on the pump and restraining the dihedral angles of PLN to retain the helical structure. Docking was guided by cross-linking data (Toyoshima et al., 2003; Chen et al., 2006; Chen et al., 2003), performing short MD runs (500 ps) with an harmonic upper wall potential applied to restrain the distances between cross-linked residues to below 5 Å. 100 such runs were performed starting from different orientations of PLN, and the resulting docked complexes were clustered according to the backbone RMSD of the proteins. The center of the most highly populated cluster was picked as the most representative structure and used to continue the modeling. Based on pairwise contacts between SERCA and PLN residues, the register of helix-helix packing was the same between our refined models and the original crystal structure. The N-terminal segment of PLN, not present in the original structure, was built as a random coil detached from SERCA. Ser16 was modeled both with and without the phosphorylation, generating two different PLN/SERCA complexes.

The structures of the PLN/SERCA and pPLN/SERCA complexes were then embedded in DMPC:POPC bilayers, mimicking the experimental conditions, and solvated with TIP3P water (Jorgensen et al., 1983). To equilibrate protein-lipid systems, we employed the two-step multiscale procedure (Stansfeld et al., 2015) in which the systems were equilibrated for 1 µs with positional restraints using the coarse-grained model MARTINI (Marrink et al., 2007) and then re-converted to full-atom descriptions using the Backward approach (Wassenaar et al., 2014). The full-atomic systems were equilibrated for 50 ns at 300 K and a further 50 ns after releasing the positional restraints. Eight equally spaced structures were extracted from the final 20 ns of sampling. These eight structures were used as starting points for ssNMR-restrained replica-averaged sampling (RAOR-MD). The final equilibrated box (of dimensions 10.8 × 10.8×15.9 nm³) contains 247 DMPC lipids, 82 POPC molecules, 39,017 TIP3P waters and 23 Na⁺ ions to neutralize the system (~174,000 atoms).

¹⁵N chemical shift (CS) and ¹⁵N-¹H dipolar coupling (DC) restraints were incorporated into the sampling using replica-averaged restrained MD, as previously described (De Simone et al., 2014; Sanz-Hernández et al., 2016) (code available at https://github.com/maximosanz/modelSSNMR), composed of eight replicas evolving simultaneously. The restraining forces were gradually incorporated during an initial equilibration phase of 20 ns, where the forces were linearly increased to a maximum of 50 J/(mol·ppm²) and 800 J/(mol·kHz²) for CSs and DCs, respectively.

In order to enhance the conformational sampling of the disordered domain Ia of PLN, we implemented a sampling based on annealing cycles, whereby the PLN N-terminus periodically binds and detaches from the SERCA surface. The experimental PRE measurements (Gustavsson et al., 2013) were incorporated in order to drive each binding event. At the beginning of each cycle, the domain Ia of PLN is fully detached from SERCA by introducing a lower wall harmonic potential that pushes PLN residues away from SERCA residues present in the cytoplasmic domains of the pump. The lower wall potential pushes PLN residues 0 to 10 at least 50 Å away from Ile140 (SERCA A domain), Thr430 (SERCA N domain), and Cys674 (SERCA P domain). Residues 11–14 were pushed at least 25 Å away from those residues and in addition the interfacial SERCA residues Thr742, Ala327, and Leu119. The detachment occurs gradually, by linearly increasing the force of the lower-wall potential to a maximum of 5 J/(mol · nm²) after 500 ps. During this time, the temperature is also linearly increased to 370 K to enhance the conformational sampling space of the disordered domain Ia. CS and DC force constants were linearly decreased to half their value to avoid instabilities during this high-energy phase of the cycle. This detachment step yields a fully detached PLN domain Ia with no SERCA contacts and enough surrounding free space to explore its unbound disordered conformational space. This step is followed by another 500 ps of sampling at 370 K, where the detached PLN domain Ia is allowed to fluctuate. Subsequently, a binding stage follows, whereby the lower-wall potential is linearly removed over 1 ns of sampling. The PRE-derived distance restraints were gradually incorporated over this time as an upper-wall potential at 30 Å with a maximum force of 5 J/(mol·nm²). During this step the force constants of CS and DC are also restored to their full value, and the temperature is linearly decreased to 300 K. After the binding phase, the PLN N-terminus adopts a SERCA-bound conformation. This step leads onto the sampling phase of the cycle, where the restraint forces and the temperature are kept constant for 2 ns. The structures sampled during these 2 ns are the ones included in the final conformational ensembles of SERCA/PLN. This annealing cycle sampling approach is illustrated in Figure 8.

We performed 25 annealing cycles per replica for both PLN and pPLN, resulting in a total simulation time of 0.8 µs for each ensemble. 10,000 equally separated structures (400 ns) in the sampling part of the cycles were extracted to perform the analyses described in the main text. All samplings were performed using a previously described version of GROMACS (Pronk et al., 2013), modified to include the CS and DC restraints (Sanz-Hernández et al., 2016). The CHARMM36 force field (Huang and MacKerell, 2013) was used. Temperature was coupled using the v-rescale algorithm (Bussi et al., 2007) and pressure was coupled at 1 bar using the semi-isotropic Berendsen method (Berendsen et al., 1984). All simulations were carried out under periodic boundary conditions. The integration timestep was set to 2 fs and the LINCS algorithm was used for constraints (Hess et al., 1997). Electrostatic interactions were accounted for using the Particle Mesh Ewald method (Darden et al., 1993). VMD (Humphrey et al., 1996) was used for contact, hydrogen bond and distance measurements; custom Python scripts for computing helical tilt (θ) and rotation (ρ) angles; and GROMACS energy tool for electrostatic interactions. For the principal component analysis (PCA), both SERCA/PLN and SERCA/pPLN ensembles were reduced to only the backbone atoms, combined, and RMS fit to the overall average coordinates prior to computing the coordinate covariance matrix using CPPTraj (Roe and Cheatham, 2013; Galindo-Murillo et al., 2015). This ensured that global rotational and translational movements were removed, and eigenvectors obtained for each ensemble could be directly compared. A PCA analysis was done with PLN (or pPLN) bound to construct videos of the PCA modes, and without PLN for clustering and the projection of X-ray structures illustrated in the main text. Based in the eigenvalues, the first two principal components accounted for 52% the overall motion out of 100 modes calculated from the ensembles. Convergence was assessed by Kullback-Leibler divergence (KLD) of the evolving 2D histogram of the projections (PC1 vs. PC2) computed using data successively truncated from time t = 0 to time t compared against the histogram determined using all timepoints:KLD(t)=∑i=0Mp(xi,t)⋅ln⁡(p(xi,t)p(xi)),where p(x_i) is the final normalized intensity of each bin i and M is the total number of bins (75²). p(x_i, t) is the normalized bin intensity at variable time t.